Oct-2012

Vacuum unit performance

Inaccurate feed characterisation and process modelling errors are major contributors to poor performance in a vacuum unit as refiners switch to heavier crude types

Scott Golden, Tony Barletta and Steve White
Process Consulting Services

Article Summary

Refinery grassroots or revamp vacuum units frequently fail to meet expected product yield, product quality or run length targets. Real performance versus design is often 3-6 wt% lower in vacuum gas oil (VGO) product yield on whole crude, with much higher VGO metals and microcarbon, and 12 to 24-month run lengths instead of 48 to 60-month targets. Lost profitability can be tens or hundreds of millions of US dollars per year.

Designers blame a low VGO yield, poor VGO quality or short run lengths on crude blend, refiner operations, equipment suppliers’ errors or numerous other perceived causes; rarely is it attributed to a lack of know-how by the designer. In this age of easy-to-use computer simulations, there is a belief that even an inexperienced engineer will be able to design a successful vacuum unit if appropriately sophisticated software is utilised. Experience proves otherwise. Low VGO product yield and poor vacuum unit reliability are becoming more common even though the global refining industry is becoming more competitive.

Today, most design engineers are experts in running process simulations and equipment models, yet almost none have validated them by comparing model output to actual measured performance. Many designers rarely have first-hand experience of the results of their work. The office-based approach presumes model results represent actual unit performance. Nowhere in the refinery can this assumption be more catastrophic than in the vacuum column simulation.

It is no surprise that basic knowledge is critical for successful vacuum unit design, including appropriate analytical tests for accurate feed characterisation, and the influence of process simulation structure is overlooked when performing a grassroots or revamp design. This article addresses fundamental feed characterisation and process simulation principles, which are part of the basic know-how required to design a vacuum unit. Faulty equipment modelling and design are also contributing factors to under-performance in a vacuum unit, but they are not part of this discussion.

Feed characterisation
Accurate feed characterisation is essential to predict VGO product yield and to meet run length targets. Without knowing the true amount of VGO product in the feed, how is it possible to predict its yield? VGO product yield is often specified as true boiling point (TBP) cutpoint. The TBP curve plots the weight or volume per cent distilled of whole crude on the X-axis against TBP temperature on the Y-axis. The TBP temperatures are produced from a standard laboratory test or series of tests. VGO product TBP cutpoint is simply the temperature taken from the curve corresponding to the percentage of total products lighter than the vacuum residue produced (see Figure 1). Starting with inaccurate TBP data or, worse, using laboratory tests such as ASTM D1160 data (that must be converted to TBP temperature) ensures unit performance will not meet expectations. Even today, many vacuum units are designed and product specification guarantees based on ASTM D1160 distillations. This laboratory test should never be used to design or monitor a vacuum unit. Unfortunately, few designers are aware of proper feed characterisation techniques, including the use of a modern, high simulated temperature distillation (HTSD) ASTM D7169 chromatographic method.

TBP curve
Crude assay TBP curves are reported as a single curve, yet they are typically created from a series of tests (see Figure 2), including ASTM D2892 and D5236 batch distillation. TBP curves are used in process simulations, in conjunction with specific gravity curves, to generate pseudo-component properties. Psuedo-component properties are utilised by generalised correlations (GS, BK10) or equations of state (PR, SRK) to generate vapour equilibrium K-values. Understanding the individual test methods and how they influence the temperatures reported on the TBP curve is important. Since the ASTM D2892 and D5236 methods use completely different methods, and operate at both atmospheric and vacuum pressures, the TBP curve generated has inherent inaccuracies that influence the reported temperatures. In practice, as crude oil gets heavier (Maya, Venezuelan crude, Canadian bitumen, Arab Heavy, Marlim, and so on), the crude assay TBP curves generated by the standard ASTM D5236 become increasingly inaccurate above 800°F (427°C), leading to poor process modelling predictions of actual vacuum unit performance.

Most crude TBP curves are generated from ASTM D2892 and ASTM D5236 test methods. TBP curves generated by these test methods have a range of 70°F (21°C) to approximately 1000°F (538°C). Maximum reported TBP temperatures vary from 960°F (516°C) to 1050°F (566°C), depending on crude type and oil stability. Rarely is it possible to operate the ASTM D5236 test above 1000°F (538°C) atmospheric equivalent temperature (AET) with heavy crudes due to poor thermal stability. Only those crude TBP curves created with an ASTM D7169 allow characterisation beyond oil thermal cracking limits.

An ASTM D2892 test uses a distillation column with 15 theoretical stages of efficiency and operates at 5/1 to 2/1 reflux ratios to fractionate the 700°F (371°C) minus portion of the whole crude into individual cuts. These samples can be further analysed for such properties as specific gravity, PIONA, cetane index, freeze, and so on. Part of the ASTM D2892 operates at atmospheric pressure and part under vacuum to keep the batch pot temperature below the oil cracking temperature. Since the ASTM D2892 column has 15 stages and uses reflux, the individual cuts have only a small distillation overlap between adjacent cuts. On the other hand, the ASTM D5236 test is a single-stage flash operated at 
2 mmHg absolute pressure or lower with no reflux to keep the pot temperature below the cracking limit. Since the ASTM D5236 does little fractionation, the reported initial and final cut temperatures are higher than they would be if the samples were well fractionated such as the ASTM D2892. This lack of fractionation causes the 700°F (371°C) plus portion of the reported TBP temperature to be higher than a well-fractionated sample with little overlap. In fact, when plotting raw test measurements, there is a step change between the D2892 and D5236 portions of the TBP curve caused by the lack of fractionation typical of the difference in D2892 and D5236 test methods. The ASTM D5236 temperature is higher.

The authors have been using TBP curves generated with high-temperature simulated distillations (HTSD) since 1994 to characterise crude oils and diesel and heavier products. The HTSD results are input to the process simulation directly as TBP by weight. More than 50 vacuum units have been field tested to confirm the validity of using HTSD (ASTM D7169) as a TBP curve to model crude and vacuum unit operation. The field work included extensive use of the HTSD to characterise the whole crude and to characterise all the product streams heavier than diesel. These field test runs also included data consistency checks, such as synthesising the whole crude from the corrected material balance and individual streams HTSD distillations and specific gravities. The synthesised TBP distillation and whole crude HTSD TBP distillation were compared to ensure data consistencies prior to input to the process model. Prior to the HTSD method being recently accepted by the ASTM, HTSDs were used by only a few refiners to characterise crude oil.